TDPL a bad idea?

[BCS]

Hello Roman,

I guess you and I have massively different opinions on this because I see 
it exactly the other way.

> Still, every time two or more people decide to answer to the same
> message, they inadvertently create new thread branches.

The linear mode prevents each reply from becoming it's own branch.

> To read new
> replies (which are usually all over the place) you need to keep
> switching contexts.

The linear model presents replies in such an order that more often than not 
sequential posts are not on the same context so I have to, more often than 
not, switch contexts for every single reply.

> I'd say the only reason it's manageable is because
> of massive embedded quotes.

In a tree mode, quoting is only *needed* indicating what part of a post you 
are replying to. In the linear mode quoting is *needed* just to give context 
if you reply to anything BUT the tail end of the thread.

What people actually *do* is a social problem, not a technical one.

> 
> With a linear thread, you can respond to several posts with just one
> message. Even better, you can respond to an entire thread without
> using quotations at all,

OK, I'll give you that one.

> so it is possible to have something akin to a
> real-life conversation with several people.
> 

One of the things I LIKE about NGs are that they DON'T have that aspect (anything 
said is implicitly in reply to only the most recent context) of real-life 
conversations. Unless you are trying to prevent long detailed dialog (and 
there are good reasons to do that in some contexts), intentionally bringing 
that in is just stupid.




As a side note, I see an interesting analogy between this conversation and 
memory models in CPUs: IIRC, the older CPUs enforced a total order on memory 
operations but as the latency on operations grew relative to the latency 
on local operations, the models started refusing to enforce a total order 
but rather only enforced a partial order. I see a similar situation here. 
Chat protocols (IRC) expect near-real-time response loops out of the user 
and can get away with a nearly pure linear model. offline protocols like 
newsgroups and e-mail expect that some users might not even see a post before 
others have cycled several times through their response loop. Those protocols 
generally present only a partial order.


--
<IXOYE><